gpt4free/tool/readme_table.py
abc c1101f9a0c ~
v0.0.2.0
2023-08-17 15:01:59 +02:00

104 lines
3.1 KiB
Python

import re
import sys
from pathlib import Path
from urllib.parse import urlparse
sys.path.append(str(Path(__file__).parent.parent))
from g4f import models, Provider
from g4f.Provider.base_provider import BaseProvider
def main():
print_providers()
print("\n", "-" * 50, "\n")
print_models()
def print_providers():
lines = [
"| Website| Provider| gpt-3.5 | gpt-4 | Streaming | Status | Auth |",
"| ------ | ------- | ------- | ----- | --------- | ------ | ---- |",
]
providers = get_providers()
for _provider in providers:
netloc = urlparse(_provider.url).netloc
website = f"[{netloc}]({_provider.url})"
provider_name = f"g4f.provider.{_provider.__name__}"
has_gpt_35 = "✔️" if _provider.supports_gpt_35_turbo else ""
has_gpt_4 = "✔️" if _provider.supports_gpt_4 else ""
stream = "✔️" if _provider.supports_stream else ""
status = (
"![Active](https://img.shields.io/badge/Active-brightgreen)"
if _provider.working
else "![Inactive](https://img.shields.io/badge/Inactive-red)"
)
auth = "✔️" if _provider.needs_auth else ""
lines.append(
f"| {website} | {provider_name} | {has_gpt_35} | {has_gpt_4} | {stream} | {status} | {auth} |"
)
print("\n".join(lines))
def get_providers() -> list[type[BaseProvider]]:
provider_names = dir(Provider)
ignore_names = [
"base_provider",
"BaseProvider",
]
provider_names = [
provider_name
for provider_name in provider_names
if not provider_name.startswith("__") and provider_name not in ignore_names
]
return [getattr(Provider, provider_name) for provider_name in provider_names]
def print_models():
base_provider_names = {
"cohere": "Cohere",
"google": "Google",
"openai": "OpenAI",
"anthropic": "Anthropic",
"replicate": "Replicate",
"huggingface": "Huggingface",
}
provider_urls = {
"Bard": "https://bard.google.com/",
"H2o": "https://www.h2o.ai/",
"Vercel": "https://sdk.vercel.ai/",
}
lines = [
"| Model | Base Provider | Provider | Website |",
"| ----- | ------------- | -------- | ------- |",
]
_models = get_models()
for model in _models:
split_name = re.split(r":|/", model.name)
name = split_name[-1]
base_provider = base_provider_names[model.base_provider]
provider_name = f"g4f.provider.{model.best_provider.__name__}"
provider_url = provider_urls[model.best_provider.__name__]
netloc = urlparse(provider_url).netloc
website = f"[{netloc}]({provider_url})"
lines.append(f"| {name} | {base_provider} | {provider_name} | {website} |")
print("\n".join(lines))
def get_models():
_models = [item[1] for item in models.__dict__.items()]
_models = [model for model in _models if type(model) is models.Model]
return [model for model in _models if model.name not in ["gpt-3.5-turbo", "gpt-4"]]
if __name__ == "__main__":
main()